After the Spill

How far should an embodied system change after unexpected experience? Research OS, Mismatch Lab, and Builder Lab by Seonhye Gu.

View the Project on GitHub higuseonhye/research-os

Study 2 — dream curriculum (public)

Experiment: EXP-SURG-002 · parallel to Paper 1 · not Paper 1 confirmatory
Status (2026-07-28): Phase 1 + Phase 2 executed · pilot archived (Tier B) · GPU closed for Study 002 label · Confirmatory → Paper 002


Start here

Doc Purpose
Phase 1 design v0.1 Frozen hypotheses · Phase 1 outcomes
Phase 2 design v0.1 Mock–Isaac alignment · H3′ · ablation v0.2
Phase 2 run protocol VESSL legs · promote · pass criteria
Selection ablation protocol Phase 1 ablation (executed)
VESSL Isaac setup v0.1 Workspace kept · Jupyter
EXP-SURG-002 README Quick start · results table
Sandbox v0.2 config Phase 2 parameters

Committed results (Phase 1)

Label Tier Summary
mock_smoke_v0.2/ B 3-seed mock · G vs D
h3_mock_isaac_v0.1/ B H3 top-k · ρ=null
selection_ablation_v0.1/ B top/bottom IR 1.0 vs 0.8
h3_mock_isaac_v0.2/ B H3 20-spec · ρ=0.15 FAIL
isaac_full_v0.1/ C Phase 1 Isaac top-k

Committed results (Phase 2 · executed 2026-07-24)

Label Tier Summary
mock_smoke_v0.4/ B v0.2 dream space · seeds 42–44
selection_ablation_v0.2/ B study1d · top 1.0 · bottom 0.3
h3_mock_isaac_v0.4/ B H3′ ρ=0.899 PASS

Public boundary

See docs/PUBLIC_BOUNDARY.md.